acl_openailab/installation.md

   1 # 1. User Quick Guide
   2 [![License](https://img.shields.io/badge/license-BSD-blue.svg)](LICENSE)
   3
   4 This Installation will help you get started to setup CaffeOnACL on RK3399 quickly.
   5
   6 # 2. Preparation
   7 ## 2.1 General dependencies installation
   8         sudo apt-get -y update
   9         sodo apt-get -y upgrade
  10         sudo apt-get install -y build-essential pkg-config automake autoconf protobuf-compiler cmake cmake-gui
  11         sudo apt-get install -y libprotobuf-dev libleveldb-dev libsnappy-dev libhdf5-serial-dev
  12         sudo apt-get install -y libatlas-base-dev libgflags-dev libgoogle-glog-dev liblmdb-dev libopenblas-dev
  13         sudo apt-get install -y libopencv-dev python-dev
  14         sudo apt-get install -y python-numpy python-scipy python-yaml python-six python-pip
  15         sudo apt-get install -y scons git
  16         sudo apt-get install -y --no-install-recommends libboost-all-dev
  17         pip install --upgrade pip
  18
  19 ## 2.2 Download source code
  20
  21         cd ~
  22
  23 #### Download "ACL" (arm_compute :[v17.09](https://github.com/ARM-software/ComputeLibrary/tree/8938bd3f40ea62ff56d6ed4e2db0a8aee34dd64a)):
  24         git clone https://github.com/ARM-software/ComputeLibrary.git
  25 #### Download "CaffeOnACL" :
  26         git clone https://github.com/OAID/CaffeOnACL.git
  27 #### Download "Googletest" :
  28         git clone https://github.com/google/googletest.git
  29
  30 # 3. Build CaffeOnACL
  31 ## 3.1 Build ACL :
  32         cd ~/ComputeLibrary
  33     aarch64-linux-gnu-gcc opencl-1.2-stubs/opencl_stubs.c -Iinclude -shared -o build/libOpenCL.so
  34         scons Werror=1 -j8 debug=0 asserts=1 neon=1 opencl=1 embed_kernels=1 os=linux arch=arm64-v8a
  35
  36 ## 3.2 Build Caffe :
  37         export ACL_ROOT=~/ComputeLibrary
  38         cd ~/CaffeOnACL
  39         cp Makefile.config.acl Makefile.config
  40         make all distribute
  41
  42 ## 3.3 Build Unit tests
  43 ##### Build the gtest libraries
  44         cd ~/googletest
  45         cmake CMakeLists.txt
  46         make
  47         sudo make install
  48
  49 ##### Build Caffe Unit tests
  50         export CAFFE_ROOT=~/CaffeOnACL
  51         cd ~/CaffeOnACL/unit_tests
  52         make clean
  53         make
  54
  55 ## 3.4 To Configure The Libraries
  56
  57         sudo cp ~/ComputeLibrary/build/libarm_compute.so /usr/lib
  58         sudo cp ~/CaffeOnACL/distribute/lib/libcaffe.so  /usr/lib
  59
  60 # 4. Run tests
  61
  62 #### 4.1 Run Caffenet
  63         cd  ~/CaffeOnACL/data/ilsvrc12
  64         sudo chmod +x get_ilsvrc_aux.sh
  65         ./get_ilsvrc_aux.sh
  66         cd ../..
  67         ./scripts/download_model_binary.py ./models/bvlc_reference_caffenet
  68         ./distribute/bin/classification.bin models/bvlc_reference_caffenet/deploy.prototxt models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel data/ilsvrc12/imagenet_mean.binaryproto data/ilsvrc12/synset_words.txt examples/images/cat.jpg
  69         output message --
  70           ---------- Prediction for examples/images/cat.jpg ----------
  71           0.3094 - "n02124075 Egyptian cat"
  72           0.1761 - "n02123159 tiger cat"
  73           0.1221 - "n02123045 tabby, tabby cat"
  74           0.1132 - "n02119022 red fox, Vulpes vulpes"
  75           0.0421 - "n02085620 Chihuahua"
  76
  77 #### 4.2 Run Unit test
  78           cd ~/CaffeOnACL/unit_tests
  79           ./test_caffe_main
  80           output message:
  81             [==========] 29 tests from 6 test cases ran. (1236 ms total) [ PASSED ] 29 tests.